Python agents

Multi-agent AI music discovery assistant that turns natural-language mood prompts into ranked Spotify recommendations and one-click playlist creation.

[ICML'26] AgentWebBench: Benchmarking Multi-Agent Coordination in Agentic Web

input: video --> output: subtitles -- AI-powered video translation and terminology management pipeline. Features multi-agent translation, RTL support, and a built-in web UI for consistent subtitles.

AI Collaboration System / 智能多 Agent 协作编排框架

Multi-agent orchestration mesh with routing, workflows, and A2A gateway

Self-Adaptive Generation Engine — Agent Development Kit with cognitive routing, evolutionary topology, formal verification, and 4-tier memory

Multi-Agent Retrieval-Augmented Generation (RAG) Customer Support System using Python, LangChain, and LangGraph.

Escaping the Median Trap in LLM Ideation — Semantic Tabu, Solution Taxonomy, and Orthogonal Insight Protocol tested across 196 solutions (8 conditions)

Framework of 43 specialized AI agents for marketing, sales, content and development automation — powered by OpenClaw & Claude AI

Application-layer protocol for AI agents serving different principals to coordinate over shared state. 📦 PyPI: https://pypi.org/project/mpac/ · Spec, Python + TypeScript reference implementations, two-machine demo.

Autonomous research agent that runs locally on a Mac via LM Studio. Plans, fetches, synthesizes, and cites — overnight, $0, with real source attribution. Plug in OpenRouter for cloud-grade synthesis when needed.

Powerful Python agent runtime with tools, MCP, Hooks, Skills, Rag, memory, sessions, reasoning, and streaming packets.

RudraX Army — 349 specialist agents, 43 model providers, 9 tactical squads and one open-source command center. Developed by Lalit Pandit.

Multi-agent system built with Claude and MCP featuring 7 specialized agents (chatbot, security researcher, business advisor, task manager, REST API), 29 shared tools across web analysis, memory, RAG, email, and communication. Includes hot-reload development, OAuth infrastructure, and remote MCP support for distributed deployments.

SkillEngine — framework-agnostic skills engine for LLM agents. Claude Code-like experience on any model.

Pi agent skills for specialized workflows and automation

An operating system for Claude Code. Skills, hooks (MCR), and a lean-context framework for smarter AI coding sessions.

Recurrsive Context Agent

Provide ready-to-use plugins, hooks, and commands to enhance Claude Code sessions with data mining, automation, and integration tools.

Offline MCP Server for Qt 4.8, Qt 5, or Qt 6 documentation

Anytime Verified Agents (AVA) is a framework for building reliable LLM-based reasoning systems that adaptively allocate computational resources under budget constraints. AVA combines adaptive search strategies, uncertainty estimation, and verification cascades to maximize reliability while respecting token, tool-call, and verification budgets.

Agentic harness for iteratively improving a local LLM's structured-output quality, scored against a frontier-model gold standard.

A privacy-first, self-hosted AI assistant with smart model routing and local LLM support (Ollama).

A vision-assisted robotic arm, with natural language voice inputs.

AI-powered ransomware threat intelligence platform. Monitor 300+ leak sites, track 24,000+ victims, and analyze threat actors with real-time dark web scraping and ML-based enrichment.

AI-powered interactive lessons from any GitHub repository

A privacy-first AI meeting assistant that lives as a floating widget on your desktop. Records microphone and system audio natively — Powered by Tauri v2, WhisperX, and LangGraph. Open-source, no bots, no meeting links — then transcribes and summarizes with your LLM of choice, fully local if you want.

VRSecretary is a production-ready reference implementation for building AI-powered assistants in virtual reality. It combines Unreal Engine 5 (VR), a FastAPI backend, local/cloud LLM backends (Ollama, IBM watsonx.ai), and Chatterbox TTS to drive an interactive VR secretary avatar. The project includes an Unreal plugin, a sample VR demo level

The ultimate MCP documentation server - 170+ curated repos, 11,000+ docs, full-text search for AI assistants

AI-powered teaching assistant that generates lesson summaries, quizzes, and AI explanations from educational content using Streamlit and LLM APIs.

AI Engineer: A multi-provider AI coding assistant with function calling, supporting DeepSeek and OpenRouter. Features include intelligent file operations, code analysis, and dynamic model switching for optimal development workflows.

Build Python data transformation pipelines with a visual IDE and AI assistant

TONE – Token-Optimized Notation Engine for LLMs is a compact, human-readable data format designed for AI and LLM contexts. It encodes JSON-compatible structures using far fewer tokens typically 40-60% less making it ideal for AI assistants, RAG systems, and conversational data pipelines.

A conscious AI-powered terminal RPG where your choices shape identity, story, and world. Guided by the MAAT principles: Harmony, Balance, Creativity, Connection, and Respect. Fully local. No cloud. No accounts. Just you and the system.

142 specialized AI agents converted to Kiro CLI JSON format — from frontend wizards to blockchain auditors

Responsible use of generative AI in assisted coding

Write in Markdown. Prompt with tags. Run from the CLI.

Comprehensive Python SDK for Venice AI - Advanced LLM, Image Generation, Audio Synthesis, and AI Assistant APIs with Full Swagger Specification Support

EGMA: Enhanced GitLab MCP API Server, comes with 143+ tools for AI-powered GitLab integration, MCP server that connects AI assistants to GitLab with intelligent context awareness, URL deduction, and specialized tools for projects, issues, MRs, pipelines, and repository operations. Works with Claude Desktop, Amazon Q, Cursor

VERÐANDI — The Norn of Becoming. AI nervous system for real-time cross-instance awareness via Unix domain sockets. Self-healing, self-correcting, conscious. The cornerstone of AGI systems. Lightweight and able to run on any Linux and Unix based systems. Use with Hermes Agent or other AI Agents.

GH-300 live demo kit. A real Python app left intentionally immature (bugs, security gaps, low coverage) across 5 structured Copilot modules. Prompt library, Skills, AGENT.md, and a reset script included.

Multi-provider web search broker for AI agents. 14 providers, budget-aware routing, content extraction — one API.

Codex용 AI 개발 규칙 문서를 생성하는 한국어 기반 룰 생성 도구

Reusable skills for AI coding agents (Claude Code, Codex, Forge) covering paper review, commit triage, GPU rentals, reference search, research logs, image-prompt composition, and more.

Lint, test, and optimize your MCP tool schemas before shipping. Catches the bugs that make LLMs pick the wrong tool.

Code for LaGR-SEQ: Language-Guided Reinforcement Learning with Sample-Efficient Querying

PAPPO is a patch-aware reinforcement learning algorithm and experimental RL framework for tool-using coding agents. It provides turn-level credit assignment, PPO training utilities, rollout extraction, critic baselines, benchmark runners, and reproducible evaluation pipelines for repository repair tasks.

A modular AI Agent framework designed for complex task orchestration and logic reasoning.

Runnable, framework-focused recipes for the Ferro Labs AI Gateway: build LangChain & LangGraph agents that route across OpenAI, Anthropic & Gemini through one endpoint — docker compose up and go.

Dynamic workflow tooling for Codex: planner/replanner orchestration, Codex CLI process swarms, broker evidence, and benchmark reports.

Open-source multi-agent AI algorithmic trading system for Interactive Brokers & MetaTrader 5 with consensus-based execution, risk management, and broker reconciliation.

A generic, batteries-included, async-first agent toolkit — configurable loop, tools, security guards, streaming, persistence, and MCP. Import as 'agentix'.

Open-source rollout, verification, and trajectory plane for agentic reinforcement learning.

The operations layer for revenue disputes. Settles chargebacks, refund requests, and failed payments in minutes - not days.

AI Agent Error Doctor: paste MCP / LangChain / CrewAI tracebacks, get root cause, exact fix, and verification steps.

An autonomous AI agent daemon — picks up tasks, executes them, validates the output, and commits results.

The lightest CLI agent for Python. No heavy deps, no complex chains. pip install, set your API key, and run AI tasks anywhere.

Codex plugin for visual fidelity gates in AI-driven UI implementation workflows.

Telegram bot from hell - Claude Code behind a Forum Topics UI

LLM-powered, channel-native Discord agent for the POAP KINGS Clash Royale clan — event-sourced reactive engine, scoped public/leadership memory, and GitHub-backed poapkings.com publishing.

A sophisticated Telegram bot for financial market analysis, featuring AI-powered SMC recommendations, economic news integration, advanced price alerts, and a multi-tier subscription system. Built with Python and Aiogram.

Full-screen live sports scoreboard dashboard skill for Claude Code

AI-powered tool for automated root cause analysis of CI/CD pipeline failures. Fetches logs from Jenkins or GitHub Actions, classifies errors using regex, semantic, and LLM-based methods, and emails structured incident reports to the right team. Supports both HTTP API and CLI modes. Self-learning for faster, smarter diagnostics.

AI Station Navigator is a modular AI workstation based on Claude Code. Install a ai skill in one prompt; execute a skill-flow in two. | AI Station Navigator 是一款基于 Claude Code 的模块化 AI 工作站。一句话安装技能，两句话运行工作流（skill-flow）。

🤖 An AI-powered DevOps agent implementing ReAct architecture to automate mobile release pipelines (Crash Check -> Perf -> Build -> Audit). Built with LangGraph & Ollama.

This is a trail of ai agent based adc design

It's a university final-exam review assistant skill.

Give ChatGPT a more complete execution environment through GitHub Actions.

A coding agent that edits real repos with workspace tools (read/write/search/run), produces JSON traces with token accounting, and renders runs on a live dashboard.

Patent Creator是一款Benchmark驱动优化的专利构建Agent

Thread-scoped runtime contracts for Codex and Claude Code, powered by MCP, local SQLite, Web UI, and UserPromptSubmit hooks

A self-evolving scaffold for autonomous web projects. 9 workflows, hourly self-evolution, self-healing pipeline, feedback learning loop. The repo is the system.

Enable AI agents to share memory across tools and avoid repeating work when switching between tasks or platforms.

A pip install defense layer that constrains AI coding agents through enforced protocols; not prompts. Pre-flight checklists, HMAC-signed session audits, shell command boundaries, 35 drift checks, and a document type system that prevents docs from becoming lies. Works with any agent: Kimi, Claude, Copilot, Codex, custom.

Mímir's Well — AI memory database with Ebbinghaus decay, self-healing, backup rotation, and full-text search

A local-first cognition core: typed, evidence-tiered, provenance-tracked memory assembled into prompts with explicit epistemics. Zero core dependencies — Python + SQLite + your own model endpoint.

A tiny AI creature that lives on an E-Ink screen, chats with a remote LLM brain, and refuses to run out of RAM. Built for the Pi Zero 2W

Less admin, more games. AI agent tools for Codecks.io — CLI-first for token efficiency, 52 MCP tools for power users. Also a standalone CLI and Python API.

A cognitive runtime that gives LLM agents persistent state, identity, and learning across turns. Memory, beliefs, drives, self-evolution, skills, and affective state — engineered scaffolding outside the model.

HermesResearch: a Telegram-native web research agent powered by Hermes via OpenRouter. Search, fetch, extract, and reply with cited markdown reports.

AI agent for querying engineering documents - upload PDFs, ask questions, detect contradictions, and run compliance checks across multiple files.

🤖 AI-powered Digital Marketing Assistant — content creation, strategy, research & analytics via Telegram. Bilingual EN/ID, Notion integration, pluggable LLM, security-hardened.

🗺️ Career trajectory exploration system — Framework that builds and assesses valid career paths through optimization, planning, and constraint satisfaction, integrating an LLM to analyze and qualitatively compare generated trajectories.

An LLM-orchestrated browser-use framework with tool-driven web control, citation/evidence capture, persistent memory, and sub-agent delegation to complete complex web tasks end-to-end.

SRE incident-investigation finite state machine for LLM agents, served over MCP by Theodosia. The agent keeps the full Grafana toolset; the FSM gates the procedure (triage, diagnose, verify, conclude) and the audit trail, not the tools.

A worked example of an agentic job-search pipeline — the Claude Code skill I ran during my 2026 job hunt.

An agentic data platform for e-commerce analytics, combining data engineering pipelines, a serving layer, and multi-agent orchestration, with a clear path toward GenAI, MLOps, DevOps, and cloud deployment.

First LLM runtime that preserves chain-of-thought under continuous user interruption. S(t+1)=f(S(t),ΔU)

Curated, annotated list of multi-agent systems resources — MARL, game theory, negotiation, LLM agent teams, MCP/A2A protocols, benchmarks, and frameworks

Ctx2Skill — 自主上下文技能发现忠实复现：arXiv:2604.27660v2

MCP-native shared memory for Claude Code agents — hybrid BM25+vector+RRF recall, cross-session continuity, cognitive scoring, knowledge graph. Self-hosted, $0/query. 30 MCP tools.

Awesome paper list for LLMs in politics, geopolitics, policymaking, strategic studies, and decision-making

A dangerous-capability eval for autonomous smart-contract vulnerability discovery: LLM agents (discover, verify, judge) scored on precision, recall, and cost against a labeled corpus and baselines.

Stripping/Compacting files to burn less tokens for coding agents.

A hybrid stellar classifier that combines deterministic astrophysical rules (Hard Computing) with a fine-tuned LLM (AstroSage-8B) to assign MK spectral types and generate natural-language stellar descriptions from Gaia DR3 data.

Sistema multi-agente de ciberseguridad ofensiva/defensiva construido con LLMs. Orquesta 6 subagentes especializados (OSINT, pentesting, hardening, forense) sobre un laboratorio Docker aislado, integrando herramientas reales (nmap, Hydra, nikto) para generar reportes estructurados de hallazgos y remediación. Proyecto de aprendizaje.

Persistent memory and session handoff for AI agents — provenance-labeled context, gated writes that can refuse, hostile test suites. Manual + six working examples.